Overview

Elemental is a framework for distributed-memory dense linear algebra that strives to be both fast and convenient. It combines ideas including: element-wise matrix distributions (Hendrickson et al.), object-oriented submatrix tracking (FLAME, van de Geijn et al.), and first-class matrix distributions (PLAPACK, van de Geijn et al.). In many instances, similar approaches to LAPACK (Anderson et al.) are taken to avoid unnecessary underflow/overflow.

So far, it has been primarily developed by Jack Poulson, but anyone is welcome to submit suggestions and/or patches. Please see the AUTHORS file for a complete list of contributors.

Installation

A thorough explanation is available in the build system section of the documentation. Developers will likely want to start by cloning the repository with

hg clone https://code.google.com/p/elemental/

Alternatively, you can download the latest tarball, which contains Elemental version 0.74.

Documentation

A working version of the documentation for the development version of Elemental is available. It is still far from complete, so please feel free to join the group and/or email the mailing list, elemental-dev@googlegroups.com.

Related publications

Three preprints are available, though most people should start with the introductory paper on Elemental. There is also an early version of a paper on reducing generalized Hermitian-definite eigenvalue problems to standard form, as well as a case study on porting Elemental to Intel's SCC chip.